Neighbor's load prediction for dynamic load balancing in a distributed computational environment

Abstract

In distributed computing environment, divisible load technique is used to speedup the completion time of a parallel task by splitting a huge task into a smaller grain size jobs where jobs can be executed remotely by other nodes. Due to the heterogeneity of computing nodes, load balancing technique is employed to distribute workload evenly across distributed nodes in order to reduce the overall response time and maximize the resource utilization. Load information plays an important role in heterogeneous computing environment to formulate a job distribution decision. In this paper, we propose a dynamic load balancing algorithm where its information policy for those nodes that have an empty neighbors' list to discover and predict other neighbors' load information. Simulation results from various scenarios show that our approach able to reduce the overall average response time of each job.

Machine Learning Techniques used for the Histopathological Image Analysis of Oral Cancer-A Review

“…It, thus, represents one of the systems' fundamental steps to classify cells regardless of context [9]. Numerous supervised [10] and unsupervised [11] machine learning algorithms have been put forward in latest years for the classification of histopathological images like support vector machines [12,13], neural networks [14], decision tree [15], fuzzy and genetic algorithms [16], k-NN [17,18], kernel PCA [19], etc. These models can be extensively used for other areas of medical science, such as medicine and clinical research [20].…”